1. Field
The present invention relates generally to a method of and apparatus for transmitting data. The invention also relates to a method of and apparatus for adaptive control of data buffering in a transmitter. The method and apparatus are well suited for use in the GPRS standard but are not limited to such an application.
2. Background
The general packet radio system (GPRS) is a packet data based communication system that has been developed for GSM networks with the aim of providing networks built to this standard with a way to handle higher data speeds and packet switched connections. GPRS can also be used in time division multiple access (TDMA) networks (IS-136). It is intended to provide a transitional path to third generation (3G) wireless data services It enables the introduction of packet switching and Internet Protocol (IP). The GPRS standard is now well defined and is currently being deployed in existing GSM-based mobile networks, in order to provide a way for GSM operators to meet the growing demand for wireless packet data services.
The GPRS standard defines a logical link control (LLC) layer which provides a logical link between a mobile station (MS) and a serving GPRS support node (SGSN). The logical link control (LLC) provides services necessary to maintain a ciphered data link between the MS and the SGSN. The logical link is maintained as the MS moves between cells serviced by the same SGSN. When the MS moves to a cell being serviced by a different SGSN the existing connection is released and a new logical link connection is established.
The logical link control (LLC) provides for acknowledged and unacknowledged point-to-point-delivery of LLC protocol data units (PDUs) between the mobile station (MS) and the serving GPRS support node (SGSN) and point to multipoint delivery of packets from the SGSN to the MS. The LLC layer also provides for detecting errors from corrupted PDUs by checking a frame check sequence (FCS) in the LLC frame format. The FCS contains the value of a cyclic redundancy check (CRC) calculation performed over a header and information fields in a frame. For the acknowledged mode of transfer, the LLC may request retransmission of the frames of data for which an acknowledgement has not been received.
Network layer protocols are intended to operate over services derived from a wide variety of sub-networks and data links. GPRS supports several network layer protocols providing protocol transparency for of the service. All functions relating to the transfer of protocol data units (PDUs) are carried out transparently by GPRS network entities. A layer known as the Sub-Network Dependant Convergence Protocol (SNDCP) provides this protocol transparency and support for a variety of network layer protocols. The SNDCP is logically situated below the network layer and above the LLC layer. It performs multiplexing of data coming from different sources before the data is sent via the logical link control (LLC) layer.
Data to be transmitted is first multiplexed by the SNDCP. The data is then segmented by the LLC layer to maximum length LLC frames. These LLC frames are segmented into radio link control (RLC) data blocks or radio link control/medium access control (RLC/MAC) control blocks, which are formatted into blocks of four successive time slots on the same physical channel.
The medium access control (MAC) layer provides capability for multiple mobile stations to share a common transmission medium. It interfaces directly with the physical layer. For the uplink (e.g. mobile station MS to a serving GPRS support node SGSN), the MAC layer plays the role of arbitrator, managing the limited physical resources among competing requestors. For the downlink, the MAC layer aids in the queuing and scheduling of access attempts and prioritizes data to be sent. Signaling data is given higher priority user data, but both are multiplexed onto the transmission medium by the MAC layer.
One problem with data transfer is that it can arrive in bursts depending on the source and/or medium from which it arrives. In one interval of time, several blocks of data may arrive in quick succession, whereas in the next interval of time only one block, or even no blocks, may arrive. Plainly, such “bursty” delivery of data is undesirable because it places overheads in terms of data management on the receiving entity. Ideally, the data should arrive at a constant rate that is as high as the receiving entity can competently handle.
One way in which “bursty” data could be handled would be to determine empirically the way in which the bursts of data generally arrive and to use a buffer large enough to maintain an essentially continuous flow of data from the source to the destination. While this approach will undoubtedly work, it is less than satisfactory because the buffer will have to be sufficiently large to hold data in the situation where a large burst of data arrives followed by a period of time when no data arrives. Most of the time a buffer of that size will be less than full and will therefore be underutilized. This is, of course, a waste of resources and is therefore undesirable.